Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
10368741 | Mechanical Systems and Signal Processing | 2017 | 14 Pages |
Abstract
An approach for seam tracking of micro gap weld whose width is less than 0.1Â mm based on magneto optical (MO) imaging technique during butt-joint laser welding of steel plates is investigated. Kalman filtering(KF) technology with radial basis function(RBF) neural network for weld detection by an MO sensor was applied to track the weld center position. Because the laser welding system process noises and the MO sensor measurement noises were colored noises, the estimation accuracy of traditional KF for seam tracking was degraded by the system model with extreme nonlinearities and could not be solved by the linear state-space model. Also, the statistics characteristics of noises could not be accurately obtained in actual welding. Thus, a RBF neural network was applied to the KF technique to compensate for the weld tracking errors. The neural network can restrain divergence filter and improve the system robustness. In comparison of traditional KF algorithm, the RBF with KF was not only more effectively in improving the weld tracking accuracy but also reduced noise disturbance. Experimental results showed that magneto optical imaging technique could be applied to detect micro gap weld accurately, which provides a novel approach for micro gap seam tracking.
Related Topics
Physical Sciences and Engineering
Computer Science
Signal Processing
Authors
Xiangdong Gao, Yuquan Chen, Deyong You, Zhenlin Xiao, Xiaohui Chen,